Anthropic researcher warns AI safety risks could be severe

An Anthropic researcher said he believes there is more than a 10% chance AI could kill all humans within the next decade, adding to growing debate over the risks of advanced systems. The comments followed viral posts from a person claiming to have quit Anthropic over concerns that AI companies are moving too quickly and not being fully transparent about the dangers. Anthropic researcher Evan Hubinger later publicly agreed that the company takes the possibility of catastrophic harm seriously. Experts interviewed in the article said the warnings reflect wider concerns about AI safety, cyberattacks, misinformation, and the lack of strong regulation. Others argued that the biggest problems are already visible today, including job displacement, weak oversight, and the rapid spread of AI into public systems and education. The story also notes that Canada’s AI policy remains incomplete. Minister Evan Solomon said the government is working on regulation to protect privacy and safety, but critics say legislation is still lagging behind the pace of development.